WP 17/01

Not for everyone: Personality, mental health, and the

use of online social networks

Dr Peter Howley and Christopher Boyce

January 2017

http://www.york.ac.uk/economics/postgrad/herc/hedg/wps/

1

Not for everyone: Personality, mental health, and the use of online social networks

Running head: Online socialising and mental health

Peter Howley¹ and Christopher Boyce²

1. Environment Department, University of York, Heslington, York, YO10 5DD, United Kingdom.

2. Stirling Management School, University of Stirling, Stirling, FK9 4LA, United Kingdom

Abstract: Much previous research has examined the relationship between online socialising and

mental health, but conclusions are mixed and often contradictory. In this present paper we unpack the

online social networking - mental health relationship by examining to what extent the relationship

between these variables is personality-specific. Consistent with the idea that communicating through

the internet is fundamentally different from face-to-face socializing, we find that on average, use of

social networking web-sites is negatively associated with mental health. However, we find that the

mental health response is dependent upon an individual’s underlying personality traits. Specifically,

individuals who are either relatively extraverted or agreeable are not substantively affected from

spending significant amounts of time on social networking web-sites. On the other hand, individuals

who are relatively more neurotic or conscientiousness are much more likely to experience substantive

reductions in their mental health from using social networking web-sites. We suggest that if the aim

of public policy is to mitigate the adverse mental health effects from excessive internet use, then one-

size-fits all measures will likely be misplaced. More generally, our research highlights the importance

of conducting differentiated analyses of internet users when examining the health effects from internet

use.

Keywords: personality traits; psychological health; internet; social interaction

JEL: I10

Corresponding author contact details:

Dr. Peter Howley

Senior Lecturer in Economics

Environment Department

University of York

Heslington, York, YO10 5DD

United Kingdom

Phone: 00 44 1904 324058

Email: peter.howley@york.ac.uk

2

Not for everyone: Personality, mental health, and the use of online social networks

Abstract: Much previous research has examined the relationship between online socialising and mental health,

but conclusions are mixed and often contradictory. In this present paper we unpack the online social networking - mental health relationship by examining to what extent the relationship between these variables is personality-specific. Consistent with the idea that communicating through the internet is fundamentally different from face-to-face socializing, we find that on average, use of social networking web-sites is negatively associated with mental health. However, we find that the mental health response is dependent upon an individual’s underlying personality traits. Specifically, individuals who are either relatively extraverted or agreeable are not substantively affected from spending significant amounts of time on social networking web-sites. On the other hand, individuals who are relatively more neurotic or conscientiousness are much more likely to experience substantive reductions in their mental health from using social networking web-sites. We suggest that if the aim of public policy is to mitigate the adverse mental health effects from excessive internet use, then one-size-fits all measures will likely be misplaced. More generally, our research highlights the importance of conducting differentiated analyses of internet users when examining the health effects from internet use.

Keywords: personality traits; psychological health; internet; social interaction

1. Introduction

There is much research to suggest that social interaction can bring about significant mental

health benefits (Cohen, 2004; Ho, 2016). Traditionally, face-to-face meetings, especially

within neighbourhoods has been an important mechanism for facilitating social interaction, as

geographical constraints meant that people commonly had few strong ties outside their locality

(Howley et al., 2015). Yet as a result of advances in transportation systems and information

technologies, many individuals have come to rely less and less on those who live in their

locality (Wellman, 2001). This leads to the question as to whether the shift towards electronic

modes of socialising has an influence on psychological outcomes. Specifically, we are

interested in whether the use of online social networks brings about the same mental health

benefits as ‘real’ social interaction? Some suggest that use of the internet is causing people to

become socially isolated and cut off from genuine social relationships, as the internet lacks

nonverbal cues important for normal social interaction (Kraut et al. 1998; Stepanikova et al.

2010). This viewpoint is based on the idea that communicating through the internet is

fundamentally different from face-to-face socializing and such ‘impoverished’ interaction may

lead to feelings of anonymity and isolation (Kraut et al., 1998). On the other hand, some

research suggests that online interaction could bring about significant benefits when it comes

3

to our psychological health by freeing people from the constraints of geography and thereby

allowing individuals to more easily pursue the types of social networks that would bring them

the most value (Robinson, 2000; Nie et al., 2002; Valentine and Holloway, 2002; Brynin and

Kraut, 2006).

Unfortunately, the empirical work in this area is somewhat mixed and often contradictory.

Kraut et al. (1998) found that greater use of the internet was associated with an increase in

individuals’ depression and loneliness. This finding was dubbed as an “Internet paradox”

because use of the internet, a technology for social contact, actually led to the reduction of

offline social ties. However, the authors subsequently found in a 3 year follow up of

respondents in their initial study that many of these negative effects dissipated (Kraut et al.,

2002). Much of the more recent work in this area has examined the association between using

the internet for social purposes and psychological health for young adults, in particular

University students. Findings from this work generally suggest a negative correlation between

time spent on social networking sites and mental health (Matsuba, 2006; Kim et al., 2006;

Whitty & McLaughlin, 2007; Kross et al., 2013). Using a large nationally representative survey

of US citizens, Stepanikova et al. (2010) found similar patterns among the general population,

i.e. as time spent on web browsing, surfing, and creating websites increased, loneliness

increased and life satisfaction decreased. Still, other research points to the potential benefits

that using the internet for socialising can have on mental health. Howard et al. (2001) and

Shklovski et al. (2004), for instance, finds a positive relationship between emailing and the

probability that people will visit friends and relatives. Robinson et al. (2000) observed that

internet users were likely to spend more time communicating face-to-face and over the phone

with family and friends. Research by Pendry and Salvatore (2015) suggests that engagement

4

with online discussion forums can have beneficial impacts on well-being and also lead to

increased engagement in offline civic action.

One potential limitation with existing research in this area, which is the focus of this study, is

that research to date has focused on ascertaining the average mental health impact of using

online social networks across society. It may, however, prove to be too simplistic to assume a

‘catch all’ average effect. Specifically, the hypothesis explored in this paper is that different

personality types will have different reasons for using the internet, and this in turn, will predict

the extent to which they are positively or negatively affected by using social networking sites

(SNSs). Consistent with the idea that the internet can cause people to become isolated from

genuine social relationships, we find that using SNSs is positively related with poor mental

health as captured by the Generalised Health Questionnaire (GHQ). We find, however, that

personality traits have an important role in moderating these effects. For instance, individuals

with relatively higher levels of neuroticism or conscientiousness are at much greater risk of

being negatively affected by use of SNSs. On the other hand, those who score relatively highly

on extraversion and agreeableness are not substantively affected from spending significant

amounts of time on SNSs. Such findings pose new questions on the link between online

socialising and mental health and highlights the importance of conducting differentiated

analyses of internet users when examining the health effects from internet use.

2. Personality traits and internet use

Personality is typically described as encompassing “the psychological component of a person

that remains from one situation to another” (Wood & Boyce, 2014) and is most typically

captured by psychologists using the influential Five Factor Model (McCrae & Costa, 2008).

The Five Factor Model consists of five over-arching traits: agreeableness, conscientiousness,

5

extraversion, neuroticism, and openness. Personality measures such as the Five Factor Model,

enable the exploration of whether an individual’s mental health response of engaging in some

activity, such as online social networking, is dependent upon an individual’s underlying

personality traits. Previous research has highlighted the important role that personality plays

not just in predicting our overall well-being (Steel et al., 2008), but also how an individual

might respond to both positive and negative life events such as unemployment (Boyce, Wood,

and Brown, 2010), widowhood (Pai and Carr, 2010), income changes (Boyce and Wood, 2010;

Boyce, Wood, and Ferguson, 2016), and retirement (Kesavayuth, Rosenman, & Zikos, 2016).

We expect personality to have a similarly (if not more so) important role in moderating the

effects from use of SNSs on mental health. This is because, personality is likely to be an

important predictor as to how individuals use and engage with the internet (Amichai-

Hamburger, 2002).

For example, neurotic individuals, who are characterised as being more prone to negative

emotions such as anxiety and anger, as well as feelings of vulnerability, have been shown to

be more likely to use the internet to feel a sense of belonging (Amiel & Sargent, 2004) and find

support for various emotional problems (Blumer & Renneberg, 2010). Although such use of

the internet can be beneficial for their mental health, neurotic individuals are also more

sensitive to negativity (Nettle, 2006) and prone to internet addiction (Kayis et al., 2016), both

of which might lead to adverse effects on mental health. Extraverts who are characterised as

being assertive, active, and outgoing generally spend less time in internet chat rooms than less

extraverted (introverted) individuals (McElroy et al., 2007) and more time using online social

networks (Correa et al., 2010). Although it has been suggested that the internet offers

advantages for introverted individuals to compensate socially (Valkenburg & Peter, 2009) it is

6

suggested that this may also result in compulsive internet use which has been shown to be

detrimental for mental health (Kayis et al., 2016).

Like extraversion and neuroticism, the role of conscientiousness in predicting how individuals

will use and engage with SNSs is not clear-cut. On the one hand, individuals who are relatively

conscientiousness in nature are less likely to experience problematic use with the internet,

because of their tendencies toward order and self-discipline (Kayis et al., 2016). Conscientious

individuals are, however, also less likely to welcome distraction (Wehrli, 2008) and therefore

overall time spent online may result in a negative impact on their well-being. Agreeable

individuals, who are characterized with high levels of trust and compliance, tend to experience

higher regret regarding online posts whilst using social media (Moore & McElroy, 2012),

whereas disagreeable individuals, who are less compliant and are less modest, may be less

likely to have favourable social interactions online (Wehrli, 2008). Finally, individuals who

score high on openness, i.e. tend to value aesthetics, creativity, and feelings, may use internet

as a tool to further their appetite for new information and insights into their understanding of

the world (Tuten & Bosnjak, 2001).

3. Data

The data used in this analysis comes from Understanding Society: the UK household

longitudinal study (UKLS). This is a comprehensive household survey that started in 2009 with

a nationally-representative stratified, clustered sample of around 50,000 adults (16+) living in

the United Kingdom. There was a supplemental ‘social networks’ module added to the main

stage questionnaire in wave 3 of the UKLS (2011-2013). This additional module asked

individuals about their use of social networking web-sites and as these questions are only asked

in this module, our analysis is limited to wave 3 of the UKLS. The indicator of mental health

7

we use is the Generalised Health Questionnaire (GHQ) which consists of a 12 item scale

designed to assess somatic symptoms, anxiety and insomnia, social dysfunction and severe

depression. It is possibly the most common assessment of mental health used in the literature

to date (Goldberg et al., 1997; Jackson, 2007). Higher scores represent relatively worse mental

health and for ease of use we refer to this indicator of mental health as psychological distress.

Within the social networks module contained in wave 3 of the UKLS, respondents were asked

two questions in relation to their use of online social networks. The first question simply asked

them to report whether they belonged to any social networking web-sites (yes/no). Overall, 45

percent of individuals report that they do belong to a social networking web-site. In the

regression analysis that follows we refer to this variable as ‘social web-site’. The second

question of interest to this study asked respondents who stated that they belonged to a social

networking web-site ‘how many hours they spend chatting or interacting with friends through

social web-sites on a normal week day, that is Monday to Friday?.’ They were given five

options ranging from none to 7 or more hours. The distribution of respondents across the

various categories can be seen in table 1 and hereafter we refer to this variable as ‘hours online’.

Social web-site and hours online formed the key explanatory variables in our regression

analysis of mental health.

Based on prior research, we include a rich set of commonly observed predictors of mental

health as control variables (see Dolan, 2008 for a review of this literature). These include

variables such as age, equivalent household income, gender, relationship status, number of

children, education and labour force status. We also included variables reflecting more

traditional means of socialising with others as additional controls such as the extent to which

individuals talk with their neighbours, whether they have close friends living nearby, and the

8

degree to which they socialise with friends in their area. To control for differences in health

which could be confounded with the extent to which individuals rely on the internet for social

interaction and their mental health, we also include a measure of individual’s general

assessment of their health in the model specification. Finally we supplement these variables

with a full set of regional controls to capture any regional differences in the extent to which

opportunities for online social networking are available to survey respondents.

To obtain a measure of personality traits, participants were asked to what extent they

agree/disagree with 25 statements beginning with the quote “I see myself as someone who”.

Each statement is classed in one of five categories: extraversion, agreeableness,

conscientiousness, neuroticism and openness. A composite score for each the Five Factor

Model personality traits is then derived by summing the scores for each of the individual

categories. To test if there are important personality-SNSs interaction effects we interact each

of our measures of personality with first our variable capturing whether a respondent is a

member of a social networking web-site (social web-site) and second with our variable

reflecting the amount of time respondents’ spend chatting online (‘hours online’). When

presenting our interaction effects, we first standardised individual’s scores on each of the five

personality traits with a mean of zero and a standard deviation of one. This aids in the

presentation of the interaction effects as it allows any unit effects to be interpreted with respect

to standard deviations.

4. Analysis

The analysis begins by assuming that mental health (H) as captured by the Generalised Health

Questionnaire is explained by a vector of socio-economic and demographic characteristics (X),

9

personality traits (P) and a dummy variable indicator capturing whether respondents are a

member of a social networking web-site (S). This yields the following explanatory model:

𝐻 = 𝑎0 + 𝛽1𝑋 + 𝛽2𝑃 + 𝛽3 𝑆 + 𝛽4(𝑆 ∗ 𝑃) + 𝜀

Support for the hypothesis that personality traits will moderate the relationship between being

a member of a social networking web-site and mental health can be obtained when the latter

two mental health determinants (S*P) is statistically significant, which would suggest that the

effect1 derived from being a member of a social networking web-site will partly depend on

individuals underlying personality traits.

Following the same procedure, albeit this time we focus solely on individuals who are a

member of a social networking web-site and add a variable (T) reflecting ‘how many hours

they spend chatting or interacting with friends through social web-sites on a normal week day,

that is Monday to Friday’ as an explanatory variable to our regression analysis such that we

end up with the following model specification:

𝐻 = 𝑎0 + 𝛽1𝑋 + 𝛽2𝑃 + 𝛽3 𝑇 + 𝛽4(𝑇 ∗ 𝑃) + 𝜀

In this specification, mental health (H) is again a function of a vector of socio-economic and

demographic characteristics (X), personality traits (P), but this time we substitute our variable

T capturing the amount of time individuals spend on social web-sites (hours online) for our

dummy indicator S. Support for the hypothesis that personality traits will moderate the mental

health effect of online social networking can again be obtained when the latter two mental

health determinants in this specification (T*P) is statistically significant.

1 Due to the mixed results from previous research we make no a priori judgement as to whether the overall association between being a member of a social networking web-site and mental health will be positive or negative

10

Under this specification we are focusing specifically on individuals who belong to a social

networking web-site when examining the relationship between time spent on SNSs (hours

online) and mental health. An alternative modelling approach would just have been to attribute

a zero value for hours online for those respondents who do not belong to a social networking

web-site. This would, in turn, have allowed us to examine the relationship between hours online

and mental health for the entire sample as opposed to just those who belong to a social

networking web-site. The main disadvantage of this approach is that it would have grouped

people who do not belong to a SNS with those who do, and it is perhaps questionable how

useful it is to uncover the mental health response from time spent on SNSs for individuals who

do not even belong to a SNS.

On the other hand, one could argue that focusing specifically on individuals who do belong to

a social networking web-site, as we do in this study, could give rise to a sample selection bias

if individuals who do not belong to a social networking web-site would have a different mental

health response from spending time using SNSs. As such, in a robustness check we tested the

sensitivity of our results discussed later to this alternative modelling approach whereby we

simply attached a zero value for the variable hours online for respondents who do not belong

to a SNS. We then ran our regression analysis on the entire sample. Our key findings discussed

later relating to the relationship between time spent chatting with friends online (hours online)

and mental health did not change when adopting this alternative modelling approach.

5. Results

11

i) Belonging to a social networking web-site

Table 2 reports the basic regression results designed to examine if there are any differences in

mental health between those who belong to a social networking web-site and those who don’t.

Specification 1 includes socio-economic, health and regional controls. The results relating to

these control variables are all along expected lines, and correspond with the results widely

documented in previous literature (see Dolan et al., 2008). We also included variables which

can be thought of as reflecting more ‘traditional’ means of socialising in the regression analysis,

and these results were also along expected lines. Specifically, individuals who regularly talk

with their neighbours, go out socially or have friends living locally are likely to have better

mental health than individuals who don’t. The key explanatory variable of interest in this

specification is the dummy variable indicating if a respondent belongs to a social networking

web-site (‘social web-site’). Under this specification, there is no significant association

between this variable and psychological distress.

One potential threat to the validity of this result is due to ‘personality induced bias’ as

personality traits are significantly correlated with mental health and internet use. As such,

Specification 2 tests the sensitivity of the results outlined under Specification 1 to the inclusion

of the Five Factor Model personality traits. The results relating to our measures of personality

are in keeping with previous research (see Steel et al., 2008). Individuals who score higher on

extraversion or conscientiousness are likely to have better mental health than individuals who

score relatively lower. We find the opposite relationship with neuroticism, i.e. neuroticism is

associated with poor mental health. Looking specifically at the findings in relation to ‘social

web-site’ we can see that there is a now a statistically significant positive association between

this variable and psychological distress. In other words, individuals who belong to a social

networking site (SNS) are likely to have significantly higher scores on our measure of

12

psychological distress (worse mental health), than individuals who don’t belong to a SNS. This

suggests that unobserved heterogeneity, i.e. personality induced bias could be significantly

biasing the results under Specification 1 and indeed much of the previous cross sectional

research which do not include personality controls in their model specification.

Next we test our hypothesis that personality traits will act to moderate the relationship between

use of SNSs and psychological distress. To examine if there are any personality-SNS

interaction effects we interact our dummy variable (social web-site) indicating if a respondent

is a member of a SNS with each of our personality measures and include these interaction

variables in our model specification. The results are presented in table 3 and we can see that

there are a number of significant interaction effects. First, looking at extraversion, we find

that the while the main effect between social web-site and psychological distress is positive,

the interaction between extraversion and social web-site is negative and statistically significant.

This suggests that extraversion attenuates downwards the negative association between being

a member of a social networking web-site and mental health. In other words, extraverts are less

likely to be negatively affected when it comes to their mental health by being a member of a

social networking web-site. On the other hand, the interaction between both neuroticism and

conscientiousness with social web-site is positive and statistically significant. This, in turn,

suggests that individuals who score relatively highly on neuroticism or conscientiousness, in

contrast to extraversion, are more likely to be negatively affected when it comes to their mental

health by being a member of a social networking web-site.

ii) Time spent on social networking web-sites

13

Next we restrict our analysis to those individuals who report that they are a member of a social

networking web-site (55% of respondents) and examine the association between the amount of

time ‘they spend chatting or interacting with friends through social web-sites on a normal week

day, that is Monday to Friday?’ (hours online) and psychological distress (see table 4).

Specification 1 includes the basic set of control variables (i.e. socio-economic, health and

regional). We can observe a clear positive relationship between hours online and psychological

distress. In other words, individuals who spend relatively more time on SNSs are likely to have

poorer mental health than individuals who spend relatively less time. Furthermore, this

relationship increases significantly with each additional hour spent on SNSs. For example, an

individual who spends one hour on SNSs will have a higher score on our measure of mental

health of approximately 0.22 units than someone who spends less than one hour. This effect

increases substantially to a 1.31 unit difference when looking at individuals who spend 7 or

more hours.

Under Specification 2 we test the sensitivity of the regression results to the inclusion of the five

personality traits. We find that the coefficients related to our key explanatory variables are

significantly smaller after adding in personality traits as additional control variables2. These

differences, like those reported between specification 1 and 2 in table 2, again support the

suggestion that failure to add in personality traits in conventional cross sectional analysis of

the relationship between use of SNSs and psychological outcomes may give rise to significant

omitted variable bias.

2 In specification 1 the point estimates were 0.22, 0.41, 0.71 and 1.31 respectively which compares to 0.17, 0.24, 0.39 and 1.01 under specification 2.

14

To test our hypothesis that personality traits will moderate the relationship between use of

SNSs and mental health, we interact our variable reflecting the amount of time individuals

spend on SNSs during a normal weekday (hours online) with each of our measures of

personality. To aid in the presentation of our results, we converted our categorical variable

reflecting the amount of hours individual’s spend on SNSs (hours online) to one continuous

variable by simply taking the midpoint of the relevant categories (see table 1) and interacted

this variable with each of our personality measures. When it comes to respondents who report

that they spend 7 or more hours on SNSs, we made the simplifying assumption that this served

as a maximum value, although in practice some respondents could at least in theory spend more

time per day on social web-sites.

There are a number of significant interaction effects which suggest that certain personality

types are more likely to be adversely affected when it comes to their mental health by using

SNSs (see table 5). These interaction effects are best illustrated in figure 1 and we can see here

that they are substantial. First, looking at conscientiousness, we can see that relatively

conscientiousness individuals (here defined as one standard deviation above the mean level)

enjoy significantly better mental health (i.e. score 0.71 lower on psychological distress) than

less conscientiousness individuals (defined as one standard deviation below the mean) when

they don’t spend any time during a normal weekday on SNSs. However, we can see that this

difference dissipates as time spent on SNSs increases, e.g. there is no significant difference in

the mental health of relatively more and less conscientiousness individuals when they spend 7

hours or more per weekday on SNSs.

When it comes to neuroticism, we can see in figure 1 that the relationship between neuroticism

and mental health is large, i.e. irrespective of online social networking use there are large level

15

differences in psychological distress between those that are one standard deviation above and

below mean levels of neuroticism. That being said, online social networking use appears to

have a more substantive negative relationship with the mental health of individuals that are

relatively more neurotic (again defined as one standard deviation above mean levels). For

example, at zero hours of use, relatively neurotic individuals (defined as one standard deviation

above the mean level) experience significantly worse mental health (i.e. score 4.65 higher on

psychological distress) than less neurotic individuals (again defined as one standard deviation

below the mean). Although as the hours of use increases, both groups experience worsening

psychological distress, this is more pronounced for those with above mean levels of

neuroticism. As such, at 7 hours use, relatively neurotic individuals score 5.40 higher on

psychological distress than less neurotic individuals. Thus the difference extends from 4.64 at

zero hours use to 5.40 at seven hours use.

Whereas individuals who score relatively highly on conscientiousness and neuroticism are

more likely to be negatively affected by time spent on SNSs, individuals who score higher on

extraversion and agreeableness are less likely to be negatively affected. For those that are high

in agreeableness or extraversion, we can see in figure 1 that more time spent using SNSs

translates to some adverse effects on mental health. However, this negative association is much

less pronounced for those with high as opposed to low levels of agreeableness or extraversion.

For instance, as can be observed in figure 1 individuals who are one standard deviation above

and below the mean level of agreeableness and extraversion respectively, have a similar score

on our psychological distress indicator when they spend one hour or less on SNSs, i.e. the

mental health of both groups are broadly comparable. However, a significant mental health gap

emerges between both these groups as the time spent on SNSs increases. Put differently,

relatively extraverted and agreeable individuals do not appear to be greatly affected from

16

spending significant amounts of time using SNSs, whereas those who are more introverted or

disagreeable appear to be negatively affected to a much more significant degree. For example,

there is an estimated 0.41 and 0.79 unit difference in the levels of psychological distress for

individual’s one standard deviation below and above the mean level of agreeableness and

extraversion respectively, when they spend 4 hours per weekday on SNSs. This increases

further to 0.77 and 1.21 for individuals who spend 7 hours or more on SNSs.

6. Conclusion

The rapid evolution of social networking web-sites have caused significant changes in the way

many people socialise and interact with each other. This leads us to the question as to the effect

of these changes on mental health. For many, the use of SNSs can have a number of positive

psychological outcomes, in that they can help individuals feel connected to others, irrespective

of where they are. Furthermore, the internet has aided the efficiency in which people can

coordinate their social activities and this may, in turn, have a double dividend in not only

reducing the stress associated with setting up social activities but also in freeing up more time

for face-to-face socialising (Stepanikova et al., 2010). While SNSs no doubt have value to

many, there is much research, however, to suggest that online social networks may not bring

about the same benefits as more traditional forms of social interaction. In fact, much recent

research documents a negative association between the use of SNSs and mental health. The

negative association between use of SNSs and mental health is thought to occur because

communicating through the internet is fundamentally different from face-to-face socializing

and such ‘impoverished’ interaction may give rise to feelings of anonymity and social isolation

(Kraut et al., 1998). There is also some recent research to suggest that social networking sites

can trigger things like FOMO (fear of missing out), and low self-esteem from comparisons

with others (Bevens et al., 2016).

17

In this study we focused on the role of personality traits in helping us to understand the

relationship between use of SNSs and mental health. An advantage of this study is that while

research to date has concentrated on the average effect from using SNSs across the entire

population, we show that the effect of using SNSs on mental health is personality-specific. In

other words, not everyone will be affected by using SNSs in the same way, and it could be

problematic to assume that they do. Specifically, individuals who can be characterised as

relatively neurotic or conscientious are much more likely to be negatively affected by using

SNSs. On the other hand, individuals who are relatively extraverted or agreeable are not

substantively affected from spending significant amounts of time on social networking web-

sites. One possible explanation for these differences is that different personality types will have

different reasons for using the internet (Amichai-Hamburger, 2002) and certain individuals

may be more prone to unhealthy internet behaviours, such as internet addiction (Kayis et al.,

2016).

In terms of future work, it seems likely that previous findings are likely to have been affected

by numerous sources of endogeneity bias. For example, given the cross sectional nature of

research in this area, many studies are likely to be affected by omitted variable bias. This study

highlighted one potential source of omitted variable bias, namely personality traits, as we find

that adding in personality measures to the regression analysis attenuated the coefficient

estimates relating to the relationship between time spent on SNSs and mental health

downwards. In addition to the presence of confounding factors, it can difficult to conclude

which variable is the cause and which is the effect. For instance, using SNSs may negatively

affect the mental health for some, but also people more at risk of poor mental health may be

18

more likely to engage with online social networks. Longitudinal and ideally experimental

designs will help disentangle the nature of these relationships.

We make no causal claims in this work relating to the effect of using SNSs on psychological

health, albeit the broad array of control variables (e.g. personality traits) used in this study is

an advantage of this relative to much of the existing research in this area. Our study has

highlighted that quite apart from issues in relation to the precise identification of the effect of

using SNSs on mental health, any findings are likely to be personality-specific. Some people

will be negatively affected by using SNSs while for others it may lead to positive outcomes. It

would be useful, therefore, for future research to aim at better understanding the heterogeneity

of impacts across society. Overall our results serve as a note of caution when using mental

health and well-being indicators for policy and/or research purposes in that assuming a ‘catch

all’ average effect and thereby not accounting for differences in how an individual might

respond to both positive and negative events may in some instances be of limited value.

References

Amichai-Hamburger Y. 2002. Internet and personality. Computers in Human Behavior 18: 1–

10.

Bevens I, Frison E. and Eggermont S. 2016. I don’t want to miss a thing”: Adolescents’ fear of

missing out and its relationship to adolescents’ social needs, Facebook use, and Facebook

related stress. Computers and Human Behaviour 64: 1-8.

Blumer T. and Renneberg B. 2010. Lebt es sich im Internet leichter? Zeitschrift Für

Psychiatrie, Psychologie Und Psychotherapie 58: 55–62.

Boyce C.J. 2010. Understanding fixed effects in human well-being. Journal of Economic

Psychology 31: 1–16.

Boyce C.J, Wood A.M. and Brown G.D.A. 2010. The dark side of conscientiousness:

Conscientious people experience greater drops in life satisfaction following unemployment.

Journal of Research in Personality 44: 535–539.

19

Boyce C.J. and Wood A.M. 2011. Personality and the marginal utility of income: Personality

interacts with increases in household income to determine life satisfaction. Journal of

Economic Behavior & Organization 78: 183–191.

Boyce C.J, Wood A.M. and Ferguson, E. 2016. Individual differences in loss aversion:

Conscientiousness predicts how life satisfaction responds to losses versus gains in income.

Personality and Social Psychology Bulletin 42: 471–84.

Brynin M, and Kraut R. 2006. Social Studies Of Domestic Information And Communication

Technologies. In R. E. Kraut, B. Malcolm, & S. Kiesler, Computers, Phones And The Internet:

Domesticating Information Technology (pp. 3-17). New York: Oxford University Press.

Cohen, S. 2004. Social relationships and health. American Psychologist 59: 676–684.

Correa T, Hinsley A.W. and De Zuniga, H G. 2010. Who interacts on the Web?: The

intersection of users’ personality and social media use. Computers in Human Behavior 26:

247–253.

Dolan P, Peasgood T. and White, M. 2008. Do we really know what makes us happy? A review

of the economic literature on the factors associated with subjective wellbeing. Journal of

Economic Psychology 29: 94 –122.

Goldberg L.R. 1993. The structure of phenotypic personality traits. American Psychologist 48:

26–34.

Goldberg D.P, Gater R, Sartorius N, Ustun T.B, Piccinelli M, Gureje O, and Rutter C. 1997.

The validity of two versions of the GHQ in the WHO study of mental illness in general health

care. Psychological Medicine 27: 191–197.

Goldberg D.P. and Williams P. 1991. A User’s Guide to the General Health Questionnaire.

Ho C.Y. 2016. Better health with more friends: the role of social capital in producing health.

Health Economics 25: 91-100.

Howard, P.E.N, Rainie L, and Jones S. 2001. Days and nights on the internet: the impact of a

diffusing technology. The American Behavioral Scientist 45(3): 383–404.

Howley P, O Neill S. and Atkinson R. 2015. Who needs good neighbors? Environment and

Planning A 47: 939-956

Jackson C. 2007. The General Health Questionnaire. Occupational Medicine 57: 79–79.

John O.P, Robins, R.W. and Pervin, L.A. 2008. Handbook of Personality, Third Edition:

Theory and Research. Guilford Press.

Kayiş A.R, Satici S.A, Yilmaz M.F, Şimşek D, Ceyhan E. and Bakioğlu F. 2016. Big five-

personality trait and internet addiction: A meta-analytic review. Computers in Human Behavior

63: 35–40.

20

Kesavayuth D, Rosenman R.E. and Zikos V. 2016. Retirement, personality, and well-being.

Economic Inquiry 54: 733–50.

Kim K, Ryu E, Chon MY, Yeun EJ, Choi SY, Seo JS. and Nam BW. 2006. Internet addiction

in Korean adolescents and its relation to depression and suicidal ideation: A questionnaire

survey. International Journal of Nursing Studies 43(2):185–192

Kraut R, Patterson M, Lundmark V, Kiesler S, Mukopadhyay T. and Scherlis W. 1998. Internet

paradox: a social technology that reduces social involvement and psychological well-being?

American Psychologist 53(9): 1017–1031

Kraut R, Kiesler S, Boneva B, Cummings J, Helgeson V, and Crawford A. 2002. Internet

paradox revisited. Journal of Social Issues, 58(1), 49–74.

Kross E, Verduyn P, Demiralp E, Park J, Lee D.S, Lin N, et al. 2013. Facebook Use Predicts

Declines in Subjective Well-Being in Young Adults. PLoS ONE 8(8).

Matsuba M.K. 2006. Searching for self and relationships online. Cyber Psychology and

Behavior 9: 275–284

McCrae R.R. and Costa P.T.J. 2008. The five-factor theory of personality. In O. P. John, R. W.

Robins, & L. A. Pervin (Eds.), Handbook of Personality, Third Edition: Theory and Research.

Guilford Press.

McElroy J. C, Hendrickson A.R, Townsend, A.M, DeMarie S.M. 2007. Dispositional factors

in internet use: personality versus cognitive style. MIS Quarterly: 809–820.

Moore K. and McElroy J.C. 2012. The influence of personality on Facebook usage, wall

postings, and regret. Computers in Human Behavior 28: 267–274.

Nettle D. 2006. The evolution of personality variation in humans and other animals. The

American Psychologist 61: 622–631.

Nie N.H, Hillygus D.S. and Erbring, L. 2002. Internet use, interpersonal relations and

sociability: a time diary study. In Barry Wellman & Caroline Haythornthwaite (Eds.), The

Internet in everyday life (pp. 216–243). Oxford: Blackwell Publishers.

Pai, M. and Carr, D. 2010. Do personality traits moderate the effect of late-life spousal loss on

psychological distress? Journal of Health and Social Behavior 51: 183–199.

Pendry L.F. and Salvatore J. 2015. Individual and social benefits of online discussion forums.

Computers in Human Behavior 50: 211-220.

Robinson J.P, Kestnbaum M, Neustadtl A. and Alvarez A. 2000. Mass media and social life

among internet users. Social Science Computer Review 18(4): 490–501.

Shklovski I, Kraut R.E. and Rainie L. 2004. The Internet and social relationships: contrasting

cross-sectional and longitudinal analyses. Journal of Computer-Mediated Communication

10(1), Article 1. Available from http:// jcmc.indiana.edu/vol10/issue1/shklovski_kraut.html A

21

Steel P, Schmidt J. and Schultz J. 2008. Refining the relationship between personality and

subjective well-being. Psychological Bulletin 134: 138-161.

Stepanikova I, Nie N.H. and He X. 2010. Time on the Internet at home, loneliness, and life

satisfaction: Evidence from panel time-diary data. Computers in Human Behavior 26: 29–338.

Tuten Ryan T. L. and Bosnjak, M. 2001. Understanding differences in Web usage: The role of

the need for cognition and the Five-Factor Model of Personality. Social Behavior and

Personality 29: 391–398.

Valentine G. and Holloway S.L. 2002. Cyberkids? Exploring children’s identities and social

networks in on-line and off-line worlds. Annals of the Association of American Geographers

92(2): 302–319.

Valkenburg, P. M. and Peter, J. 2009. Social Consequences of the Internet for Adolescents: A

Decade of Research. Current Directions in Psychological Science 18: 1–5.

Wood A.M. and Boyce C.J. 2014. Personality, an overview. In A. C. Michalos (Ed.),

Encyclopedia of Quality of Life Research (pp. 4773–4775). Dordrecht, Netherlands: Springer.

Wehrli S. 2008. Personality on Social Network Sites: An Application of the Five Factor Model

(ETH Zurich Sociology Working Paper No. 7). ETH Zurich, Chair of Sociology. Retrieved

from https://ideas.repec.org/p/ets/wpaper/7.html

Wellman B. 2001. Computer networks as social networks. Science 293: 2031-2034.

Wellman B, Quan, A, Witte J. and Hampton K. 2001. Does the Internet increase, decrease or

supplement social capital? Social networks, participation, and community commitment.

American Behavioral Scientist 45: 436–455.

Whitty M.T. and McLaughlin D. 2007. Online recreation: The relationship between loneliness,

internet self-efficacy and the use of the internet for entertainment purposes. Computers in

Human Behavior 23(3): 1435-1446.

22

Table 1: Hours spent interacting with friends through social websites 1 Freq. Percent

None 3,269 15.79

Less than an hour 11,144 53.83

1-3 hours 4,926 23.79

4-6 hours 922 4.45

7 or more hours 443 2.14

Total 20,704 100

2

23

Table 2: Factors related with psychological distress

Specification 1 Specification 2 Coef. Std. Err. Coef Std. Err

Equivalent household income -0.35 *** 0.04 -0.27 *** 0.04

Age 0.00 0.00 0.03 *** 0.00

Female 1.20 *** 0.05 0.33 *** 0.05

In a relationship -0.28 *** 0.06 -0.33 *** 0.05

Number of children 0.02 0.03 0.09 *** 0.03

Had a degree qualification -0.10 * 0.06 -0.03 0.05

Self-employed - employed is the reference category -0.05 0.10 0.11 0.09

Unemployed 2.18 *** 0.12 1.75 *** 0.11

Retired -0.47 *** 0.10 -0.60 *** 0.09

Familycare 0.78 *** 0.11 0.42 *** 0.10

Training -0.06 0.12 -0.27 *** 0.11

Disabled 6.31 *** 0.15 4.64 *** 0.14

Other -0.29 1.03 -0.81 0.92

Talk regularly with neighbours -0.35 *** 0.03 -0.17 *** 0.03

Local friends mean a lot - 0.55 *** 0.03 -0.40 *** 0.03

Go out socially -1.36 *** 0.08 -0.93 *** 0.07

Completely satisfied with health -1.36 *** 0.08 -1.68 *** 0.08

Social web-site 0.05 0.06 0.12 ** 0.06

Regional control variables unreported for parsimony

Openness 0.02 ** 0.02

Agreeableness 0.00 ** 0.02

Extraversion -0.09 ** 0.02

Neuroticism 1.59 ** 0.02

Conscientiousness -0.32 ** 0.02

N 40,452 40,386

R² 0.14 0.30

*** statistically significant at 1% level, ** 5 % level, * 10 % level

24

Table 3: Personality interactions

Personality – social website interactions Coef. Std. Err.

Openness to experience 0.052 0.049

Agreeableness -0.053 0.049

Extraversion -0.125 *** 0.049

Neuroticism 0.179 *** 0.047

Conscientiousness 0.129*** 0.050

*** statistically significant at 1% level, ** 5 % level, * 10 % level

*Note: We exclude the results relating to the control variables from the personality interactions

column both in this table and table 5 as they are similar to those reported in table 2 and 4.

25

Table 4: Factors related with psychological distress

Specification 1 Specification 2 Coef. Std. Err. Coef Std. Err

Equivalent household income -0.42 *** 0.06 -0.31 *** 0.06

Age 0.02 *** 0.00 0.05 *** 0.00

Female 1.19 *** 0.08 0.35 *** 0.08

In a relationship -0.37 *** 0.09 -0.44 *** 0.08

Number of children 0.01 0.04 0.06 * 0.04

Had a degree qualification -0.07 0.08 -0.01 0.08

Self-employed - employed is the reference category -0.07 0.15 0.08 0.14

Unemployed 2.00 *** 0.17 1.66 *** 0.15

Retired -1.17 *** 0.21 -1.21 *** 0.19

Familycare 0.90 *** 0.16 0.54 *** 0.14

Training 0.08 0.14 -0.12 0.13

Disabled 7.02 *** 0.25 5.25 *** 0.23

Other -0.76 1.46 -1.28 1.31

Talk regularly with neighbours -0.31 *** 0.05 - 0.13 *** 0.04

Local friends mean a lot -0.60 *** 0.05 - 0.45 *** 0.04

Go out socially -1.16 *** 0.13 -0.68 *** 0.11

Completely satisfied with health -2.95 *** 0.12 -1.75 *** 0.11

Hours online (None is the reference category)

Less than one hour 0.22 ** 0.11 0.17 * 0.10

One to three hours 0.41 *** 0.13 0.24 ** 0.12

Four to six hours 0.71 *** 0.21 0.39 ** 0.19

Seven or more hours 1.31 *** 0.29 1.01 *** 0.26

Regional control variables unreported for parsimony

Openness 0.05 0.03

Agreeableness -0.03 0.04

Extraversion -0.14 *** 0.03

Neuroticism 1.65 *** 0.03

Conscientiousness -0.26 *** 0.04

N 19,246 19,237

R² 0.14 0.30

*** statistically significant at 1% level, ** 5 % level, * 10 % level

26

Table 5: Personality interactions

Personality – hours online interactions Coef. Std. Err.

Openness to experience 0.010 0.026

Agreeableness -0.061 *** 0.024

Extraversion -0.071 *** 0.027

Neuroticism 0.054 ** 0.025

Conscientiousness 0.052 ** 0.025

*** statistically significant at 1% level, ** 5 % level, * 10 % level

27

Figure 1: Psychological distress from hours spent online moderated by personality

i) Agreeableness ii) Extrvaersion

ii) Neuroticism iv) Conscientiousness